diff options
author | Lars Ellenberg <lars@linbit.com> | 2016-06-14 00:26:31 +0200 |
---|---|---|
committer | Jens Axboe <axboe@fb.com> | 2016-06-13 21:43:07 -0600 |
commit | 9104d31a759fbade8505f38f92f4dde719957826 (patch) | |
tree | ad71e3ed1cbb0a096fae43d8a48db469b8ff4761 /drivers/block/drbd/drbd_protocol.h | |
parent | 60bac040129720837770d0d852cf71dccbc69f61 (diff) | |
download | linux-stable-9104d31a759fbade8505f38f92f4dde719957826.tar.gz linux-stable-9104d31a759fbade8505f38f92f4dde719957826.tar.bz2 linux-stable-9104d31a759fbade8505f38f92f4dde719957826.zip |
drbd: introduce WRITE_SAME support
We will support WRITE_SAME, if
* all peers support WRITE_SAME (both in kernel and DRBD version),
* all peer devices support WRITE_SAME
* logical_block_size is identical on all peers.
We may at some point introduce a fallback on the receiving side
for devices/kernels that do not support WRITE_SAME,
by open-coding a submit loop. But not yet.
Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Diffstat (limited to 'drivers/block/drbd/drbd_protocol.h')
-rw-r--r-- | drivers/block/drbd/drbd_protocol.h | 74 |
1 files changed, 70 insertions, 4 deletions
diff --git a/drivers/block/drbd/drbd_protocol.h b/drivers/block/drbd/drbd_protocol.h index 95ca4586af59..4d296800f706 100644 --- a/drivers/block/drbd/drbd_protocol.h +++ b/drivers/block/drbd/drbd_protocol.h @@ -64,6 +64,11 @@ enum drbd_packet { P_RS_THIN_REQ = 0x32, /* Request a block for resync or reply P_RS_DEALLOCATED */ P_RS_DEALLOCATED = 0x33, /* Contains only zeros on sync source node */ + /* REQ_WRITE_SAME. + * On a receiving side without REQ_WRITE_SAME, + * we may fall back to an opencoded loop instead. */ + P_WSAME = 0x34, + P_MAY_IGNORE = 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */ P_MAX_OPT_CMD = 0x101, @@ -110,8 +115,11 @@ struct p_header100 { u32 pad; } __packed; -/* these defines must not be changed without changing the protocol version */ -#define DP_HARDBARRIER 1 /* depricated */ +/* These defines must not be changed without changing the protocol version. + * New defines may only be introduced together with protocol version bump or + * new protocol feature flags. + */ +#define DP_HARDBARRIER 1 /* no longer used */ #define DP_RW_SYNC 2 /* equals REQ_SYNC */ #define DP_MAY_SET_IN_SYNC 4 #define DP_UNPLUG 8 /* not used anymore */ @@ -120,6 +128,7 @@ struct p_header100 { #define DP_DISCARD 64 /* equals REQ_DISCARD */ #define DP_SEND_RECEIVE_ACK 128 /* This is a proto B write request */ #define DP_SEND_WRITE_ACK 256 /* This is a proto C write request */ +#define DP_WSAME 512 /* equiv. REQ_WRITE_SAME */ struct p_data { u64 sector; /* 64 bits sector number */ @@ -133,6 +142,11 @@ struct p_trim { u32 size; /* == bio->bi_size */ } __packed; +struct p_wsame { + struct p_data p_data; + u32 size; /* == bio->bi_size */ +} __packed; + /* * commands which share a struct: * p_block_ack: @@ -164,8 +178,23 @@ struct p_block_req { * ReportParams */ -#define FF_TRIM 1 -#define FF_THIN_RESYNC 2 +/* supports TRIM/DISCARD on the "wire" protocol */ +#define DRBD_FF_TRIM 1 + +/* Detect all-zeros during resync, and rather TRIM/UNMAP/DISCARD those blocks + * instead of fully allocate a supposedly thin volume on initial resync */ +#define DRBD_FF_THIN_RESYNC 2 + +/* supports REQ_WRITE_SAME on the "wire" protocol. + * Note: this flag is overloaded, + * its presence also + * - indicates support for 128 MiB "batch bios", + * max discard size of 128 MiB + * instead of 4M before that. + * - indicates that we exchange additional settings in p_sizes + * drbd_send_sizes()/receive_sizes() + */ +#define DRBD_FF_WSAME 4 struct p_connection_features { u32 protocol_min; @@ -240,6 +269,40 @@ struct p_rs_uuid { u64 uuid; } __packed; +/* optional queue_limits if (agreed_features & DRBD_FF_WSAME) + * see also struct queue_limits, as of late 2015 */ +struct o_qlim { + /* we don't need it yet, but we may as well communicate it now */ + u32 physical_block_size; + + /* so the original in struct queue_limits is unsigned short, + * but I'd have to put in padding anyways. */ + u32 logical_block_size; + + /* One incoming bio becomes one DRBD request, + * which may be translated to several bio on the receiving side. + * We don't need to communicate chunk/boundary/segment ... limits. + */ + + /* various IO hints may be useful with "diskless client" setups */ + u32 alignment_offset; + u32 io_min; + u32 io_opt; + + /* We may need to communicate integrity stuff at some point, + * but let's not get ahead of ourselves. */ + + /* Backend discard capabilities. + * Receiving side uses "blkdev_issue_discard()", no need to communicate + * more specifics. If the backend cannot do discards, the DRBD peer + * may fall back to blkdev_issue_zeroout(). + */ + u8 discard_enabled; + u8 discard_zeroes_data; + u8 write_same_capable; + u8 _pad; +} __packed; + struct p_sizes { u64 d_size; /* size of disk */ u64 u_size; /* user requested size */ @@ -247,6 +310,9 @@ struct p_sizes { u32 max_bio_size; /* Maximal size of a BIO */ u16 queue_order_type; /* not yet implemented in DRBD*/ u16 dds_flags; /* use enum dds_flags here. */ + + /* optional queue_limits if (agreed_features & DRBD_FF_WSAME) */ + struct o_qlim qlim[0]; } __packed; struct p_state { |